Computer vision is the interdisciplinary scientific field that develops theories and methods allowing computers to extract high-level information from digital images or videos. From an engineering perspective it seeks to automate perceptual tasks normally performed by the human visual system. Generally, vision is difficult because it is an inverse problem, where only insufficient information is available about the objects of interest in the image data. Physics-based mathematical and statistical models as well as machine-learning methods are used to assist in the task. Current real-world applications are wide-ranging, and include optical character recognition, machine inspection, retail object recognition, 3D model building, remote sensing, medical imaging, autonomous driving, motion capture, surveillance, face recognition, biometrics, and many others. This course provides an introduction to fundamental concepts and an opportunity to develop a real-world application of computer vision.

The course aims to give students a broad understanding of both classical and modern computer vision theories and methods, as well as practical skills in implementing and developing computer vision algorithms and applications.

In particular, the course will teach students about the formation process and characteristics of digital images, and the workings of techniques for image filtering and enhancement, feature extraction and representation, object detection and pattern recognition, image segmentation and classification, motion estimation and object tracking, and a wide range of applications. In addition to classical computer vision methods, students will also learn about modern machine learning and deep learning approaches for these tasks, and acquire practical skills in using them to solve real-world computer vision problems.

As computer vision is a broad, interdisciplinary field with many possible applications, the course intends to lay the theoretical (computer science) as well as practical (computer engineering) foundation to address future computer vision challenges. Also, since solutions to big challenges typically require not only individual but also team efforts, the course includes both labs and a group project to help develop the necessary skills through practical experience complementing the knowledge acquired in the lectures.

Tools such as GitHub Copilot, ChatGPT, Google Bard, and other tools based on large (language) models or other generative artificial intelligence techniques, look likely to become heavily used by programmers. However, you need a good understanding of the language you are coding in and the systems involved before you can effectively use these tools. Using these tools to generate code instead of writing the code yourself will hinder your learning. Therefore, in this course, you are not permitted to submit code generated by such tools. Submitting code generated by such tools will be treated as plagiarism and penalised accordingly.

Achieve a composite mark of at least 50 out of 100.

The course schedule includes 2 x 2-hour lectures per week during all weeks of term except flexibility week. In addition, a 1-hour consultation session is scheduled per week starting in the second week of term, where tutors are available to explain the labs (in the first weeks of the course) and the group project (in later weeks) and answer any questions about these.

All course materials will be provided online. There is no need to buy a book. In the lectures we will be referring to various resources for further reading, many of which are freely available online:

• Richard Szeliski. Computer Vision: Algorithms and Applications. Second Edition, Springer, 2021.
• Dana H. Ballard and Christopher M. Brown. Computer Vision. Prentice Hall, 1982.
• Ian Goodfellow, Yoshua Bengio, Aaron Courville. Deep Learning. MIT Press, 2016.
• David A. Forsyth and Jean Ponce. Computer Vision: A Modern Approach. Prentice Hall, 2011.
• Simon J. D. Prince. Computer Vision: Models, Learning and Inference. Cambridge University Press, 2012.

Other resources of interest (available from the library or perhaps online as well) include:

• Linda G. Shapiro and George C. Stockman. Computer Vision. Prentice Hall, 2001.
• Rafael C. Gonzalez and Richard E. Woods. Digital Image Processing. Addison Wesley, 2008.
• Milan Sonka, Vaclav Hlavac, Roger Boyle. Image Processing, Analysis and Machine Vision. Chapman and Hall, 2007.
• Richard O. Duda, Peter E. Hart, David G. Stork. Pattern Classification. John Wiley and Sons, 2000.
• Gérard Medioni and Sing Bing Kang. Emerging Topics in Computer Vision. Prentice Hall, 2005.

And furthermore we will cite many scientific articles that you may consult for more detailed information.

This course is evaluated every term using the myExperience system. The feedback provided by the students is carefully analysed by the convenor and lecturers, and points of improvement are taken on board where possible. Based on the student feedback in past terms, the following changes will be introduced this term:

• Part of the lectures will be in person but other course components will remain online.
• Lectures will be livestreamed and provide an opportunity to interact with the lecturer.
• Labs will contain more introductory material and explanation in consultation sessions.
• Consultations will remain online but will be scheduled separately from the lectures.
• Two lectures will be scheduled on the topic of deep learning for computer vision.
• Feedback on the labs will be provided faster so it can be taken into account for later labs.
• Several administrative changes will streamline project group formation and assessment.